REVIEW 3 major objections 5 minor 4 references
On the Density of Prime Imbalances in the Unit Interval
T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Prime imbalances (p−q)/(p+q) are dense in the open unit interval, the paper proves with an explicit search algorithm.
desk verdict The result is true but standard, and the proof as written has two load-bearing errors, so the paper should be desk-rejected rather than refereed. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the map $(p,q) \mapsto \frac{p-q}{p+q}$, which is monotone in the ratio $r = p/q$. The paper inverts the target by setting $r = \frac{1+t}{1-t}$, so that the exact equality holds when $p = r q$. The targeted construction then looks for a prime $p$ within a relative distance $\delta$ of $r q$; the error in the normalized difference is bounded by $3\delta$, so choosing $\delta < \varepsilon/3$ suffices. Bertrand's postulate supplies the prime inside such an interval, while the elementary bound $\pi(x) \ge x/(2\log x)$ is used to lower-bound the number of prime pairs in the direct-search argument.
What would settle it
Compute all values $\frac{p-q}{p+q}$ for primes $p > q$ up to some large bound, say $10^6$, and check every subinterval of length $0.01$ inside $(0,1)$. If any such subinterval contains no value, the paper's direct-search guarantee is refuted; more decisively, exhibiting any open interval in $(0,1)$ that contains no element of $S$ at all would refute the density theorem itself.
Extended reading notes
Core claim
The paper's central claim is that $S$ is dense in $(0,1)$, made precise by Theorem 3: for every $t \in (0,1)$ and every $\varepsilon > 0$, there exist primes $p > q$ with $\left|\frac{p-q}{p+q} - t\right| < \varepsilon$. The construction proceeds by mapping the target $t$ to the required ratio $r = \frac{1+t}{1-t}$ of the two primes, then searching for primes near $r q$. For targets not handled by this targeted search, a direct search over all prime pairs up to a bound $N_0$ is invoked, with $N_0$ chosen so that the average spacing of the values $\frac{p-q}{p+q}$ is below the tolerance. The paper also states complexity bounds for the search size, namely $N = O\left(\frac{1}{\varepsilon}\log\frac{1}{\varepsilon}\right)$ in the worst case.
Load-bearing premise
The direct-search part of the proof assumes that because the number of prime pairs with $p,q \le N$ grows like $N^2/\log^2 N$ and those pairs lie in the bounded interval $(0,1)$, their values must become $\varepsilon$-dense once the average spacing is below $\varepsilon$; this inference fails if the values cluster rather than spread evenly.
Editorial extensions
If this is right
- For every $t \in (0,1)$ and $\varepsilon > 0$, there exist primes $p > q$ with $\left|\frac{p-q}{p+q} - t\right| < \varepsilon$, and the paper's algorithm finds them in finite time.
- The density of $S$ in $(0,1)$ is equivalent, under the map $t \mapsto \frac{1+t}{1-t}$, to the statement that ratios $p/q$ of primes are dense in $(1,\infty)$.
- The complexity bound $N = O\left(\frac{1}{\varepsilon}\log\frac{1}{\varepsilon}\right)$ gives a quantitative recipe: smaller tolerances require larger primes only logarithmically in $1/\varepsilon$ beyond the linear factor.
- The construction is uniform over the whole open interval, with no exceptional behavior near $0$ or $1$.
Reading between the lines
- Editorial extension: A fully rigorous version of the direct-search step would need to show that every subinterval of $(0,1)$ actually contains many values of $\frac{p-q}{p+q}$, not merely that the average spacing is small; the paper's average-spacing formulation is a weaker condition.
- Beyond the paper: The same normalization applies to other increasing sequences, such as powers or polynomial values, and the question of whether their normalized differences are dense becomes a testable analogue that depends on the sequence's distribution in short intervals.
- Testable extension: One could run the paper's algorithm for a grid of small $t$ and $\varepsilon$ values (for example, $t = k/100$, $\varepsilon = 0.01$) and record the largest prime used, which would provide empirical data on how the complexity bound compares with actual behavior.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims to prove that the set S = {(p−q)/(p+q) : p>q are primes} is dense in (0,1). The proof strategy is constructive: Lemma 2 attempts a targeted construction around q·(1+t)/(1−t), Lemma 3 attempts a direct search with a bound N0 on the primes, and Theorem 3 combines the two. Theorem 4 gives complexity bounds for the search. The paper also includes illustrative computational examples.
Significance. If the claim were established by elementary methods, it would be a neat observation about the distribution of prime ratios. However, the manuscript does not provide a valid proof: both supporting lemmas contain load-bearing errors. The central claim may be true, but the arguments given do not establish it. The computational examples are not a substitute for a proof, and the claimed quantitative bounds are unsupported. I can identify no machine-checked proofs or reproducible code that would offset these gaps.
major comments (3)
- [Lemma 2, Eqs. (2)–(5)] The algebraic identity in Eq. (3) is incorrect. Expanding the numerator of (2) gives 2q(p−qr), not (p−qr)(q+qr)+qr(q−p). Consequently Eq. (4) and the bound (5) do not follow. In addition, the assertion that the interval [qr(1−δ), qr(1+δ)] contains a prime for sufficiently small δ is not a consequence of Bertrand's postulate; it requires a prime-gap theorem guaranteeing a prime in intervals of length x^θ with θ<1 and explicit constants, which is neither stated nor proved. Thus Lemma 2 does not establish the targeted-construction case.
- [Lemma 3] The inference from the number of prime pairs to ε-density is logically invalid. A lower bound on the number of pairs (p,q) does not imply a lower bound on the number of distinct values of (p−q)/(p+q), since many pairs can map to the same or clustered values. Moreover, small average spacing between consecutive values does not preclude a large uncovered interval: for example, all points in [0,1/log^2 N] have average spacing much smaller than ε but leave most of (0,1) uncovered. The displayed bound N0 = (2 log N0/ε)^{1/2} is self-referential and is not solved for N0. Therefore the direct-search guarantee is unproved, and Theorem 3's second case collapses.
- [Theorem 4, Eq. (7)] The complexity bound is internally inconsistent. Eq. (7) states N = O((1/ε) log(1/ε)), but the proof concludes N = O((1/√ε) log(1/ε)) for the direct search and N = O(1/ε) for the targeted method; no single bound is established. Since both underlying lemmas are invalid, the quantitative claims are unsupported.
minor comments (5)
- [Lemma 1] The proof of Lemma 1 is a citation without derivation; if the bound π(x) ≥ x/(2 log x) for x ≥ 25 is used, please state the precise reference or give a short argument.
- [Theorem 3, Step 1] The phrase 't is bounded away from 1' is not quantified. Provide an explicit threshold (e.g., t ≤ 1 − η for some absolute η) for when the targeted construction applies.
- [Section 4, Remark 1] The search in Example 1 uses 'try nearby primes' without specifying the search rule, so it does not illustrate a well-defined algorithm.
- [Theorem 4 proof] The phrase 'Solving this implicit equation yields ...' is not shown; the self-referential equation N = sqrt(log^2 N / ε) has no demonstrated solution for N, and the subsequent O(·) estimate is not justified.
- [Throughout] There are several typos and formatting issues, e.g., 'qare primes' in the abstract and missing parentheses in Eq. (7). These should be corrected.
Circularity Check
No significant circularity: the paper attempts a direct construction and does not assume the density conclusion or fit the target values.
full rationale
The paper's central claim is that S is dense in (0,1), and the proof tries to establish this by searching for prime pairs approximating a target ratio. The stated inputs — Bertrand's postulate and elementary prime-counting estimates — do not themselves contain the density conclusion. Lemma 3's step from an average-spacing estimate to ε-density is mathematically invalid, because a set of many points can still leave large gaps; but that is a logical gap, not circularity, since average spacing is not defined as equivalent to ε-density. Lemma 2 relies on a strong, unproved assertion about primes in short intervals and contains algebraic errors, but those are unsupported inputs and incorrect manipulations, not a reduction of the theorem to itself. The self-referential display 'N0 = ((2 log N0/ε)^{1/2})' is a sloppy implicit bound rather than a circular derivation: the existence of a large N satisfying log^2 N/N^2 < ε is a standard consequence of the asymptotic behavior of that ratio. There are no fitted parameters, no predictions of fitted data, and no load-bearing self-citations. The proof may be seriously defective, but it is not circular in the sense of assuming its conclusion or renaming its inputs as outputs.
Assumptions & free parameters
assumptions (4)
- standard math Bertrand's Postulate
- standard math Prime counting estimate π(x) ≥ x/(2 log x) for x ≥ 25
- domain assumption For any fixed r>0 and sufficiently small δ>0, the interval [r q - δ r q, r q + δ r q] contains a prime for all sufficiently large prime q
- ad hoc to paper A set of N points in (0,1) with average spacing d is ε-dense whenever d < ε
Cite this review
Pith. "Pith review of On the Density of Prime Imbalances in the Unit Interval." pith.science (2026). https://pith.science/paper/DCUEIMTA
@misc{pith2026250612063,
author = {Pith},
title = {Pith review of: On the Density of Prime Imbalances in the Unit Interval},
year = {2026},
howpublished = {\url{https://pith.science/paper/DCUEIMTA}},
note = {Machine review of arXiv:2506.12063}
}
abstract
We prove that the set of normalized differences between primes, defined as $S = \{(p-q)/(p+q) : p > q \text{ are primes}\}$, is dense in the open unit interval $(0,1)$. Our proof provides an explicit construction algorithm with quantitative bounds, relying on elementary results from prime number theory including Bertrand's postulate and explicit bounds on prime gaps in long intervals.
Reference graph
Works this paper leans on
-
[1]
J. Bertrand, M´ emoire sur le nombre de valeurs que peut prendre une fonction quand on y permute les lettres qu’elle renferme , Journal de l’ ´Ecole Polytechnique 17 (1845), 123–140
-
[2]
G.H. Hardy and E.M. Wright, An Introduction to the Theory of Numbers , 6th edition, Oxford University Press, 2008
work page 2008
-
[3]
Apostol, Introduction to Analytic Number Theory , Springer-Verlag, 1976
T.M. Apostol, Introduction to Analytic Number Theory , Springer-Verlag, 1976
work page 1976
-
[4]
Landau, Handbuch der Lehre von der Verteilung der Primzahlen , Teubner, Leipzig, 1909
E. Landau, Handbuch der Lehre von der Verteilung der Primzahlen , Teubner, Leipzig, 1909. 5
work page 1909
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.